A mobile energy storage device charging control method and device

Abstract

A charging control method and device for mobile energy storage equipment, the method extracts feature vectors from each node in a graph model; a graph attention network model uses the feature vectors extracted from each node to aggregate information of neighboring nodes by learning attention weights between nodes to predict the user's charging needs and the available power of the charging station; the parameters of the graph attention network model of each local node are obtained, the parameters of the graph attention network model of each local node are aggregated and updated to optimize the global model, and the optimized global model parameters are distributed to each local node; the global optimal solution is found according to the prediction results of the optimized global model, and the optimal charging station allocation and charging time plan for each mobile energy storage object are obtained. The present invention can improve the dynamic adaptability and accuracy of the charging strategy, protect user privacy, and solve the problems of low computing efficiency and accuracy and poor security in traditional technologies.

Description

Technical Field

[0001] The present invention belongs to the technical field of mobile energy storage objects, and in particular relates to a charging control method and device for mobile energy storage equipment. Background Art

[0002] At present, with the popularization of mobile energy storage objects such as electric vehicles, the pressure on the power grid is increasing, especially during peak hours, the power demand and the effective allocation of charging piles have become urgent problems to be solved.

[0003] Traditional charging strategies are mostly based on static planning, which makes it difficult to adapt to the real-time changing grid load and the dynamic charging needs of mobile energy storage objects. In addition, these methods often ignore the interaction between mobile energy storage object users and their impact on the grid state, resulting in low charging efficiency and uneven grid load distribution. Existing charging scheduling methods lack an effective mechanism to guide the charging period of mobile energy storage objects to match the peak production capacity of renewable energy, and are prone to leaking user privacy during the processing and analysis process, and are not very secure. Summary of the invention

[0004] To this end, the present invention provides a charging control method and device for a mobile energy storage device, which can not only improve the dynamic adaptability and accuracy of the charging strategy, but also protect user privacy and solve the problems of low computing efficiency and accuracy and poor security existing in traditional technologies.

[0005] In order to achieve the above object, the present invention provides the following technical solution: a charging control method for a mobile energy storage device, comprising:

[0006] Acquire interaction data between a mobile energy storage object user and a charging station, and use the interaction data to construct a graph model, wherein nodes in the graph model represent mobile energy storage object users or charging stations, and edges in the graph model represent interaction relationships between users or between users and charging stations;

[0007] Extracting feature vectors from each node in the graph model, wherein the feature vectors of the mobile energy storage object user nodes in the graph model include historical charging behaviors and charging preference parameters; the feature vectors of the charging station nodes in the graph model include power station location, capacity, current status, and historical load parameters;

[0008] Constructing a graph attention network model, which uses the feature vector extracted from each node and aggregates the information of neighboring nodes by learning the attention weights between nodes to predict the user's charging needs and the available power of the charging station;

[0009] A charging station or several charging stations in a geographical area is regarded as a local node, each of the local nodes trains a graph attention network model through local data, obtains parameters of the graph attention network model of each local node, aggregates and updates the parameters of the graph attention network model of each local node to optimize the global model, and distributes the optimized global model parameters to each local node;

[0010] The global optimal solution is found based on the prediction results of the optimized global model, and the optimal charging station allocation and charging time plan for each mobile energy storage object are obtained.

[0011] As a preferred solution for the charging control method of mobile energy storage equipment, in the process of constructing the graph attention network model, for each node Updated feature vector The calculation formula is:

[0012] ;

[0013] In the formula, Representation Node The set of neighbor nodes of represents the attention coefficient; is the trainable weight matrix, Representation Node The characteristic vector of represents a nonlinear activation function;

[0014] Attention coefficient The calculation formula is:

[0015] ;

[0016] In the formula, Represents a trainable weight vector used to calculate any two nodes i and j The attention coefficient between them; LeakyReLU(·) is a linear rectification function used to introduce nonlinear features; is the normalization factor, ensuring that the sum of the normalization coefficients of all neighbor nodes j of node i is 1; Indicates a connection operation;

[0017] In the process of building the graph attention network model, the loss function is defined The graph attention network model is trained by gradient descent method, and the loss function The calculation formula is:

[0018] ;

[0019] In the formula, represents the number of nodes in the graph attention network model, that is, the total number of mobile energy storage device users and charging stations, For Node The real label data, It is the feature representation of the model after being updated by the attention mechanism.

[0020] As a preferred solution for the charging control method of mobile energy storage equipment, a charging station or several charging stations in a geographical area is used as a local node. Each local node trains a graph attention network model through local data. The calculation formula is:

[0021] ;

[0022] In the formula, Indicates Wheel at the node Model parameters on ; Indicates Wheel at the node The model parameters on is the hyperparameter learning rate, Representation Node The gradient of the loss function calculated on the local data.

[0023] As a preferred solution of the mobile energy storage device charging control method, before obtaining the parameters of the graph attention network model of each local node, it also includes:

[0024] For each local node, differential privacy noise is added to the model update of the local node, and the expression formula is:

[0025] ;

[0026] In the formula, Indicates Wheel at the node Model parameters on ; surface shows the added differential privacy noise; Indicates Wheel at the node Model parameters after adding differential privacy noise.

[0027] As a preferred solution for the charging control method of mobile energy storage equipment, the parameters of the graph attention network model of each local node are aggregated and updated to optimize the expression formula of the global model:

[0028] ;

[0029] In the formula, Represents the total number of nodes participating in federated learning; It represents the parameters of the global model after t+1 rounds of iterations, which are obtained by aggregating the noisy model parameters uploaded by all participating nodes and are used to update the global model on the central server.

[0030] As the optimal solution for the charging control method of mobile energy storage equipment, the global optimal solution is found according to the prediction results of the optimized global model, and the optimal charging station allocation and charging time planning process for each mobile energy storage object is obtained:

[0031] Define a set of mobile energy storage objects In a group of charging stations Each mobile energy storage object has a charging requirement and arrival time ; Each charging station Maximum available power and a set of charging piles; the goal is to assign the most suitable charging station and charging pile to each mobile energy storage object, and plan the charging time of each mobile energy storage object to minimize the total charging time and cost while considering the grid load balance;

[0032] Define a binary decision vector If the mobile energy storage object Assigned to ,but ,otherwise ; Charging time of each mobile energy storage object By charging demand and charging station charging rate The calculations show that:

[0033] ;

[0034] The optimization goal is to minimize the total charging cost , including charging time, charging waiting time and optimizing grid load balance. Assuming that the charging cost is proportional to the charging time, the objective function is expressed as:

[0035] ;

[0036] In the formula, and is a hyperparameter balancing coefficient used to adjust the relative importance of charging time and grid load; Representative charging station The load is the total charging power of all mobile energy storage objects assigned to the station; is the average load of all charging stations, which is used to evaluate the degree of load balancing;

[0037] The constraints of the global model include:

[0038] Each mobile energy storage object is assigned to only one charging station:

[0039] ;

[0040] Charging station capacity limitations:

[0041] ;

[0042] Restrictions on arrival time, charging time and charging station operation time of mobile energy storage objects:

[0043] ;

[0044] The optimal charging station allocation and charging time plan for each mobile energy storage object are obtained by solving the problem.

[0045] As a preferred solution for the charging control method of mobile energy storage equipment, after obtaining the optimal charging station allocation and charging time plan for each mobile energy storage object, it also includes evaluating the execution effect of the current charging scheduling strategy and adjusting the current charging scheduling strategy according to the evaluation results.

[0046] As a preferred solution for the charging control method of mobile energy storage equipment, in the process of evaluating the execution effect of the current charging scheduling strategy, a charging waiting time evaluation model, a power grid load balance evaluation model and an energy cost evaluation model are constructed;

[0047] The expression formula of the charging waiting time evaluation model is:

[0048] ;

[0049] In the formula, represents the average waiting time, Represents the total number of mobile energy storage objects charged during the evaluation period, Representative The waiting time of a vehicle;

[0050] The expression formula of the power grid load balance evaluation model is:

[0051] ;

[0052] In the formula, L var represents the variance of the grid load; Representative charging station The load is the total charging power of all mobile energy storage objects assigned to the station; is the average load of all charging stations, which is used to evaluate the degree of load balancing; M is the total number of charging stations;

[0053] The expression formula of the energy cost assessment model is:

[0054] ;

[0055] In the formula, is the total energy cost, It is The unit energy cost of charging a vehicle, It is Vehicle charging needs.

[0056] The present invention further provides a mobile energy storage device charging control device, which adopts the above-mentioned mobile energy storage device charging control method, comprising:

[0057] A graph model building module, used to obtain interaction data between mobile energy storage object users and charging stations, and use the interaction data to build a graph model, wherein nodes in the graph model represent mobile energy storage object users or charging stations, and edges in the graph model represent interaction relationships between users, or between users and charging stations;

[0058] A feature vector extraction module is used to extract feature vectors from each node in the graph model, wherein the feature vectors of the mobile energy storage object user nodes in the graph model include historical charging behaviors and charging preference parameters; the feature vectors of the charging station nodes in the graph model include power station location, capacity, current status, and historical load parameters;

[0059] A graph attention network construction module is used to construct a graph attention network model. The graph attention network model uses the feature vector extracted from each node to aggregate the information of neighboring nodes by learning the attention weights between nodes to predict the user's charging needs and the available power of the charging station;

[0060] A federated learning optimization module, used to treat a charging station or several charging stations in a geographical area as a local node, each of the local nodes trains a graph attention network model through local data, obtains parameters of the graph attention network model of each local node, aggregates and updates the parameters of the graph attention network model of each local node to optimize the global model, and distributes the optimized global model parameters to each local node;

[0061] The resource scheduling and allocation module is used to find the global optimal solution based on the optimized global model prediction results, and obtain the optimal charging station allocation and charging time plan for each mobile energy storage object.

[0062] As a preferred solution of the mobile energy storage device charging control device, it also includes:

[0063] The strategy evaluation and adjustment module is used to evaluate the execution effect of the current charging scheduling strategy and adjust the current charging scheduling strategy according to the evaluation results.

[0064] The present invention has the following advantages: obtaining the interaction data between the mobile energy storage object user and the charging station, and using the interaction data to construct a graph model, wherein the nodes in the graph model represent the mobile energy storage object user or the charging station, and the edges in the graph model represent the interaction relationship between users, or between users and charging stations; extracting feature vectors from each node in the graph model, wherein the feature vectors of the mobile energy storage object user nodes in the graph model include historical charging behaviors and charging preference parameters; the feature vectors of the charging station nodes in the graph model include the power station location, capacity, current status, and historical load parameters; constructing a graph attention network model, wherein the graph attention network model uses the feature vectors extracted from each node to The attention weights between learning nodes aggregate the information of neighboring nodes to predict the user's charging needs and the available power of the charging station; a charging station or several charging stations in a geographical area is taken as a local node, each of the local nodes trains a graph attention network model through local data, obtains the parameters of the graph attention network model of each local node, aggregates and updates the parameters of the graph attention network model of each local node to optimize the global model, and distributes the optimized global model parameters to each local node; finds the global optimal solution according to the prediction results of the optimized global model, and obtains the optimal charging station allocation and charging time plan for each mobile energy storage object. The present invention can more efficiently utilize charging station resources and reduce waiting time in queues by predicting charging needs and measuring resource availability; ensure that user data is fully protected while providing personalized charging services to avoid the risk of data leakage; reduce the load on the power grid during peak hours, promote the stable operation of the power grid and effectively improve the renewable energy absorption capacity; quickly respond to user charging needs, guide user diversion, and improve service quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0066] Figure 1 A schematic diagram of a flow chart of a mobile energy storage device charging control method provided in an embodiment of the present invention;

[0067] Figure 2 A schematic diagram of an application scenario of the mobile energy storage device charging control method provided in an embodiment of the present invention;

[0068] Figure 3 This is a schematic diagram of the architecture of a mobile energy storage device charging control device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0069] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0070] In recent years, the development of smart grids has provided new technical means to improve charging efficiency and optimize grid management. In this context, some charging strategies based on real-time data analysis have emerged, aiming to match the power supply capacity of the grid by dynamically adjusting the allocation of charging piles and charging rates.

[0071] However, these methods still face challenges in computational efficiency and accuracy when dealing with the complex interactions of large-scale mobile energy storage objects and charging pile networks. In addition, as the proportion of distributed energy and renewable energy in the power grid gradually increases, how to promote the efficient use of these energy sources while ensuring the stability of the power grid has also become an important issue. Existing charging scheduling methods often fail to fully take this into account, lack an effective mechanism to guide the charging period of mobile energy storage objects to match the peak production capacity of renewable energy, and are prone to leaking user privacy during the processing and analysis process, and are not very secure.

[0072] In view of this, an embodiment of the present invention proposes a new charging pile control method. By introducing a graph attention neural network (GAT), the interactions between users of mobile energy storage objects and the complex relationship between them and the state of the power grid can be effectively learned. Utilizing the powerful graph structure data processing capability of GAT, efficient modeling and optimization of large-scale mobile energy storage object charging networks can be achieved, thereby improving the dynamic adaptability and accuracy of charging strategies. In addition, combined with federated learning technology, continuous optimization and updating of the model can be achieved while protecting user privacy, further improving the performance of the entire system. Thereby, not only can the utilization rate of charging piles and the balance of power grid load be optimized, but also the efficient use of renewable energy can be promoted, providing an innovative solution for realizing a more intelligent and sustainable mobile energy storage object charging network. The following is the specific content of an embodiment of the present invention.

[0073] Example 1

[0074] See also Figure 1In Embodiment 1 of the present invention, a method for controlling charging of a mobile energy storage device is provided, comprising the following steps:

[0075] S1. Obtain interaction data between a mobile energy storage object user and a charging station, and use the interaction data to construct a graph model, wherein nodes in the graph model represent mobile energy storage object users or charging stations, and edges in the graph model represent interaction relationships between users, or between users and charging stations;

[0076] S2. Extract feature vectors from each node in the graph model. The feature vectors of the mobile energy storage object user nodes in the graph model include historical charging behaviors and charging preference parameters; the feature vectors of the charging station nodes in the graph model include power station location, capacity, current status, and historical load parameters.

[0077] S3. Constructing a graph attention network model, wherein the graph attention network model uses the feature vector extracted from each node and aggregates the information of neighboring nodes by learning the attention weights between nodes to predict the user's charging needs and the available power of the charging station;

[0078] S4. Take a charging station or several charging stations in a geographical area as a local node, train each local node with a graph attention network model through local data, obtain parameters of the graph attention network model of each local node, aggregate and update the parameters of the graph attention network model of each local node to optimize the global model, and distribute the optimized global model parameters to each local node;

[0079] S5. Find the global optimal solution based on the prediction results of the optimized global model, and obtain the optimal charging station allocation and charging time plan for each mobile energy storage object.

[0080] In this embodiment, in step S1, a graph model is constructed to model the interaction between the mobile energy storage object user and the charging station. Each node in the graph model represents a mobile energy storage object user or charging station, and the edge represents the potential interaction between users or between users and charging stations. Using the graph attention neural network (GAT), the dependency between nodes can be learned to predict each user's charging needs, arrival time, and available power of the power station.

[0081] Among them, in the process of obtaining the interactive data between the mobile energy storage object user and the charging station, the collected data is divided into three categories based on the data type: user data, charging station data, and power grid data. User data mainly refers to the user's charging needs, historical orders, charging preferences, etc. obtained through the application or charging interface. Charging station data mainly refers to the location, capacity, current status, load and historical load of the charging station obtained through the pile management system or station management system. Power grid data mainly refers to the power grid load situation, electricity price information, user incentive policies, etc. obtained through power companies or power operators.

[0082] The data preprocessing stage includes outlier detection and data standardization. Since data preprocessing is performed at the local node and the data volume is not large, clustering is used to identify outliers, such as records of abnormally high charging demand and charging power exceeding the limit, identify outliers and remove them. The data is standardized to facilitate subsequent processing. Thus, data preprocessing provides high-quality input for the subsequent construction of graph models and training of graph attention networks, which helps to improve the accuracy and generalization of the model.

[0083] In this embodiment, in step S2, a feature vector is extracted for each node in the graph model. For users, the feature vector mainly includes parameters such as historical charging behavior and charging preference; for charging stations, the feature vector mainly includes parameters such as power station location, capacity, current status, historical load, etc.

[0084] In this embodiment, in step S3, the graph attention network model architecture includes an input layer, three graph attention layers, and an output layer. Based on the feature vector obtained in the preprocessing stage as input, a three-layer graph attention layer is designed. Each layer aggregates neighbor node information through the attention mechanism, and captures the complex dependencies between nodes by learning the weights of each node and its neighbor nodes (i.e., the attention coefficient). The last layer is the output layer, which is used to output the predicted target.

[0085] In this embodiment, in step S3, during the process of constructing the graph attention network model, the graph attention network aggregates the information of neighboring nodes by learning the attention weights between nodes. Updated feature vector The calculation formula is:

[0086] ;

[0087] In the formula, Representation Node The set of neighbor nodes of Represents the attention coefficient, that is, the node j For Node i the importance of is the trainable weight matrix, Representation Node The characteristic vector of represents a nonlinear activation function; the expression is .

[0088] Among them, the attention coefficient The calculation formula is:

[0089] ;

[0090] In the formula, Represents a trainable weight vector used to calculate any two nodes i and j The attention coefficient between them; LeakyReLU(·) is a linear rectification function used to introduce nonlinear features; is the normalization factor, ensuring that the node i All neighbor nodes of j The sum of the normalized coefficients is 1;

[0091] Among them, in the process of building the graph attention network model, the loss function is defined The graph attention network model is trained by the gradient descent method, and the loss function is represented by the mean square error (MSE). The calculation formula is:

[0092] ;

[0093] In the formula, represents the number of nodes in the graph attention network model, that is, the total number of mobile energy storage device users and charging stations, For Node The real label data, It is the feature representation of the model after being updated by the attention mechanism.

[0094] In this embodiment, in step S4, during the local model training process, a charging station or a plurality of charging stations in a geographical area is taken as a local node, and each local node trains a graph attention network model through local data. The calculation formula is:

[0095] ;

[0096] In the formula, Indicates Wheel at the node Model parameters on ; Indicates Wheel at the node The model parameters on is the hyperparameter learning rate, Representation Node The gradient of the loss function calculated on the local data.

[0097] Before obtaining the parameters of the graph attention network model of each local node, the method further includes:

[0098] For each local node, differential privacy noise is added to the model update of the local node, and the expression formula is:

[0099] ;

[0100] In the formula, Indicates Wheel at the node Model parameters on ; surface shows the added differential privacy noise; Indicates Wheel at the node Model parameters after adding differential privacy noise.

[0101] See also Figure 2 In practice, the child nodes upload the model parameters to the central server, and the central server aggregates and updates the parameters of the graph attention network model of each local node to optimize the global model. The expression formula is:

[0102] ;

[0103] In the formula, Represents the total number of nodes participating in federated learning; It represents the parameters of the global model after t+1 rounds of iterations, which are obtained by aggregating the noisy model parameters uploaded by all participating nodes and are used to update the global model on the central server.

[0104] The uploaded data only includes the parameter updates of the model, but not the original input data, to enhance privacy protection. The central server distributes the aggregated model to all local nodes before conducting the next round of training.

[0105] Specifically, after the graph attention network model training is completed, each node independently updates its model parameters based on local data. Before each node uploads the model update to the central server, differential privacy technology is applied to add noise to the local model parameters to ensure data privacy while minimizing the impact on model performance. The noisy model parameters are uploaded to the central server. The uploaded data only includes the parameter update of the model, but not the original input data, to enhance privacy protection. The central server aggregates the noisy model update parameters uploaded by all nodes and uses the federated averaging algorithm to update the global model. The updated global model is then distributed to each node for the next round of local training. After each round of training, the performance of the global model is evaluated and the performance indicators are fed back to each child node. The node continues to adjust its own hyperparameters based on the performance of the global model.

[0106] In this embodiment, in step S5, based on the prediction results of the optimized global model, the mobile energy storage object is intelligently scheduled to the most suitable charging station, and resources are dynamically allocated. The scheduling problem is modeled using mixed integer linear programming (MILP), and the global optimal solution is found in combination with the prediction results of the optimized global model, that is, while meeting user needs, optimizing resource utilization efficiency and minimizing charging time.

[0107] Specifically, the global optimal solution is found based on the prediction results of the optimized global model, and the optimal charging station allocation and charging time planning process for each mobile energy storage object is obtained as follows:

[0108] Scheduling problem definition: Define a set of mobile energy storage objects In a group of charging stations Each mobile energy storage object has a charging requirement and arrival time ; Each charging station Maximum available power and a set of charging piles; the goal is to assign the most suitable charging station and charging pile to each mobile energy storage object, and plan the charging time of each mobile energy storage object to minimize the total charging time and cost while considering the grid load balance;

[0109] Model Details: Defining a Binary Decision Vector If the mobile energy storage object Assigned to ,but ,otherwise ; Charging time of each mobile energy storage object By charging demand and charging station charging rate The calculations show that:

[0110] ;

[0111] The optimization goal is to minimize the total charging cost , including charging time, charging waiting time and optimizing grid load balance. Assuming that the charging cost is proportional to the charging time, the objective function is expressed as:

[0112] ;

[0113] In the formula, and is a hyperparameter balancing coefficient used to adjust the relative importance of charging time and grid load; Representative charging station The load is the total charging power of all mobile energy storage objects assigned to the station; is the average load of all charging stations, which is used to evaluate the degree of load balancing;

[0114] Among them, the constraints of the global model include:

[0115] Each mobile energy storage object is assigned to only one charging station:

[0116] ;

[0117] Charging station capacity limitations:

[0118] ;

[0119] Restrictions on arrival time, charging time and charging station operation time of mobile energy storage objects:

[0120] ;

[0121] By solving the above MILP problem, the optimal charging station allocation and charging time plan for each mobile energy storage object are obtained, so as to predict the idle state of each charging station and the charging demand of the mobile energy storage object in the future, and determine the low-load period according to the real-time load of the power grid. Then, based on the dynamic pricing of the power grid, the personalized scheduling strategy is determined according to the power grid incentive policy and the user's charging preference, historical behavior and urgency of charging demand.

[0122] In a possible embodiment, after obtaining the optimal charging station allocation and charging time plan for each mobile energy storage object, step S6 is also included to evaluate the execution effect of the current charging scheduling strategy, and adjust the current charging scheduling strategy according to the evaluation result to continuously optimize the charging resource allocation and utilization efficiency.

[0123] Among them, performance indicators are defined to evaluate the charging scheduling effect, mainly including charging waiting time, user satisfaction, grid load conditions, etc., and relevant performance indicator data are collected, analyzed and processed, and the calculated performance indicator values ​​are compared with the preset target values ​​to evaluate the execution effect of the strategy at this time.

[0124] Specifically, in the process of evaluating the execution effect of the current charging scheduling strategy, a charging waiting time evaluation model, a power grid load balance evaluation model and an energy cost evaluation model are constructed;

[0125] The expression formula of the charging waiting time evaluation model is:

[0126] ;

[0127] In the formula, represents the average waiting time, Represents the total number of mobile energy storage objects charged during the evaluation period, Representative The waiting time of a vehicle;

[0128] The expression formula of the power grid load balance evaluation model is:

[0129] ;

[0130] In the formula, L var represents the variance of the grid load; Representative charging station The load is the total charging power of all mobile energy storage objects assigned to the station; is the average load of all charging stations, which is used to evaluate the degree of load balancing; M is the total number of charging stations;

[0131] The expression formula of the energy cost assessment model is:

[0132] ;

[0133] In the formula, is the total energy cost, It is The unit energy cost of charging a vehicle, It is Vehicle charging needs.

[0134] Among them, if certain performance indicators fail to meet expectations, analyze the reasons, diagnose the specific links and causes of the problems, adjust the charging resource allocation plan on this basis, and continuously monitor the implementation effect during actual operation.

[0135] In summary, the embodiment of the present invention obtains the interaction data between the mobile energy storage object user and the charging station, and uses the interaction data to build a graph model, wherein the nodes in the graph model represent the mobile energy storage object user or the charging station, and the edges in the graph model represent the interaction relationship between users, or between users and charging stations; a feature vector is extracted for each node in the graph model, and the feature vector of the mobile energy storage object user node in the graph model includes historical charging behavior and charging preference parameters; the feature vector of the charging station node in the graph model includes the power station location, capacity, current status, and historical load parameters; a graph attention network model is built, and the graph attention network model uses the feature vector extracted from each node to The attention weights between learning nodes aggregate the information of neighboring nodes to predict the user's charging needs and the available power of the charging station; a charging station or several charging stations in a geographical area is taken as a local node, each of the local nodes trains a graph attention network model through local data, obtains the parameters of the graph attention network model of each local node, aggregates and updates the parameters of the graph attention network model of each local node to optimize the global model, and distributes the optimized global model parameters to each local node; finds the global optimal solution according to the prediction results of the optimized global model, and obtains the optimal charging station allocation and charging time plan for each mobile energy storage object. The present invention can more efficiently utilize charging station resources and reduce waiting time in queues by predicting charging needs and measuring resource availability; ensure that user data is fully protected while providing personalized charging services to avoid the risk of data leakage; reduce the load on the power grid during peak hours, promote the stable operation of the power grid and effectively improve the renewable energy absorption capacity; quickly respond to user charging needs, guide user diversion, and improve service quality.

[0136] It should be noted that the method of the embodiment of the present disclosure can be performed by a single device, such as a computer or a server. The method of the present embodiment can also be applied in a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiment of the present disclosure, and the multiple devices will interact with each other to complete the described method.

[0137] It should be noted that the above describes some embodiments of the present disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the above embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0138] Example 2

[0139] See also Figure 3 Embodiment 2 of the present invention further provides a mobile energy storage device charging control device, which adopts a mobile energy storage device charging control method of the above embodiment 1, including:

[0140] A graph model building module 001 is used to obtain interaction data between a mobile energy storage object user and a charging station, and use the interaction data to build a graph model, wherein the nodes in the graph model represent the mobile energy storage object user or the charging station, and the edges in the graph model represent the interaction relationship between users, or between users and charging stations;

[0141] The feature vector extraction module 002 is used to extract feature vectors from each node in the graph model. The feature vectors of the mobile energy storage object user nodes in the graph model include historical charging behaviors and charging preference parameters; the feature vectors of the charging station nodes in the graph model include power station location, capacity, current status, and historical load parameters;

[0142] A graph attention network construction module 003 is used to construct a graph attention network model, which uses the feature vector extracted from each node to aggregate the information of neighboring nodes by learning the attention weights between nodes to predict the user's charging needs and the available power of the charging station;

[0143] The federated learning optimization module 004 is used to treat a charging station or a plurality of charging stations in a geographical area as a local node, each of the local nodes trains a graph attention network model through local data, obtains parameters of the graph attention network model of each local node, aggregates and updates the parameters of the graph attention network model of each local node to optimize the global model, and distributes the optimized global model parameters to each local node;

[0144] The resource scheduling and allocation module 005 is used to find the global optimal solution according to the optimized global model prediction results, and obtain the optimal charging station allocation and charging time plan for each mobile energy storage object.

[0145] In this embodiment, in the graph attention network construction module 003, for each node Updated feature vector The calculation formula is:

[0146] ;

[0147] In the formula, Representation Node The set of neighbor nodes of represents the attention coefficient; is the trainable weight matrix, Representation Node The characteristic vector of represents a nonlinear activation function;

[0148] Attention coefficient The calculation formula is:

[0149] ;

[0150] In the formula, Represents a trainable weight vector used to calculate any two nodes i and j The attention coefficient between them; LeakyReLU(·) is a linear rectification function used to introduce nonlinear features; is the normalization factor, ensuring that the sum of the normalization coefficients of all neighbor nodes j of node i is 1; Indicates a connection operation;

[0151] In the graph attention network construction module 003, the loss function is defined The graph attention network model is trained by gradient descent method, and the loss function The calculation formula is:

[0152] ;

[0153] In the formula, represents the number of nodes in the graph attention network model, that is, the total number of mobile energy storage device users and charging stations, For Node The real label data, It is the feature representation of the model after being updated by the attention mechanism.

[0154] In this embodiment, in the federated learning optimization module 004, a charging station or a plurality of charging stations in a geographical area is taken as a local node, and each local node trains a graph attention network model through local data. The calculation formula is:

[0155] ;

[0156] In the formula, Indicates Wheel at the node Model parameters on ; Indicates Wheel at the node The model parameters on is the hyperparameter learning rate, Representation Node The gradient of the loss function calculated on the local data.

[0157] In this embodiment, in the federated learning optimization module 004, differential privacy noise is added to the model update of each local node, and the expression formula is:

[0158] ;

[0159] In the formula, Indicates Wheel at the node Model parameters on ; surface shows the added differential privacy noise; Indicates Wheel at the node Model parameters after adding differential privacy noise.

[0160] In this embodiment, in the federated learning optimization module 004, the expression formula for aggregating and updating the parameters of the graph attention network model of each local node to optimize the global model is:

[0161] ;

[0162] In the formula, Represents the total number of nodes participating in federated learning; It represents the parameters of the global model after t+1 rounds of iterations, which are obtained by aggregating the noisy model parameters uploaded by all participating nodes and are used to update the global model on the central server.

[0163] In this embodiment, in the resource scheduling and allocation module 005, the global optimal solution is found according to the optimized global model prediction result, and the optimal charging station allocation and charging time planning process for each mobile energy storage object is obtained:

[0164] Define a set of mobile energy storage objects In a group of charging stations Each mobile energy storage object has a charging requirement and arrival time ; Each charging station Maximum available power and a set of charging piles; the goal is to assign the most suitable charging station and charging pile to each mobile energy storage object, and plan the charging time of each mobile energy storage object to minimize the total charging time and cost while considering the grid load balance;

[0165] Define a binary decision vector If the mobile energy storage object Assigned to ,but ,otherwise ; Charging time of each mobile energy storage object By charging demand and charging station charging rate The calculations show that:

[0166] ;

[0167] The optimization goal is to minimize the total charging cost , including charging time, charging waiting time and optimizing grid load balance. Assuming that the charging cost is proportional to the charging time, the objective function is expressed as:

[0168] ;

[0169] In the formula, and is a hyperparameter balancing coefficient used to adjust the relative importance of charging time and grid load; Representative charging station The load is the total charging power of all mobile energy storage objects assigned to the station; is the average load of all charging stations, which is used to evaluate the degree of load balancing;

[0170] The constraints of the global model include:

[0171] Each mobile energy storage object is assigned to only one charging station:

[0172] ;

[0173] Charging station capacity limitations:

[0174] ;

[0175] Restrictions on arrival time, charging time and charging station operation time of mobile energy storage objects:

[0176] ;

[0177] The optimal charging station allocation and charging time plan for each mobile energy storage object are obtained by solving the problem.

[0178] This embodiment also includes: a strategy evaluation and adjustment module 006, which is used to evaluate the execution effect of the current charging scheduling strategy and adjust the current charging scheduling strategy according to the evaluation result.

[0179] In this embodiment, in the strategy evaluation and adjustment module 006, in the process of evaluating the execution effect of the current charging scheduling strategy, a charging waiting time evaluation model, a power grid load balance evaluation model and an energy cost evaluation model are constructed;

[0180] The expression formula of the charging waiting time evaluation model is:

[0181] ;

[0182] In the formula, represents the average waiting time, Represents the total number of mobile energy storage objects charged during the evaluation period, Representative The waiting time of a vehicle;

[0183] The expression formula of the power grid load balance evaluation model is:

[0184] ;

[0185] In the formula, L var represents the variance of the grid load; Representative charging station The load is the total charging power of all mobile energy storage objects assigned to the station; is the average load of all charging stations, which is used to evaluate the degree of load balancing; M is the total number of charging stations;

[0186] The expression formula of the energy cost assessment model is:

[0187] ;

[0188] In the formula, is the total energy cost, It is The unit energy cost of charging a vehicle, It is Vehicle charging needs.

[0189] It should be noted that the information interaction, execution process and other contents between the modules of the above-mentioned device are based on the same concept as the method embodiment in Example 1 of the present application, and the technical effects they bring are the same as those of the method embodiment of the present application. For specific contents, please refer to the description in the method embodiment shown in the previous part of the present application, and will not be repeated here.

[0190] Example 3

[0191] Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium, in which a program code of a mobile energy storage device charging control method is stored, and the program code includes instructions for executing the mobile energy storage device charging control method of embodiment 1 or any possible implementation thereof.

[0192] Computer-readable storage media can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0193] Example 4

[0194] Embodiment 4 of the present invention provides an electronic device, including: a memory and a processor;

[0195] The processor and the memory communicate with each other via a bus; the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the mobile energy storage device charging control method of embodiment 1 or any possible implementation thereof.

[0196] Specifically, the processor can be implemented by hardware or by software. When implemented by hardware, the processor can be a logic circuit, an integrated circuit, etc.; when implemented by software, the processor can be a general-purpose processor implemented by reading software codes stored in a memory. The memory can be integrated in the processor or can be located outside the processor and exist independently.

[0197] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center.

[0198] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0199] Although the present invention has been described in detail above by general description and specific embodiments, it is obvious to those skilled in the art that some modifications or improvements can be made on the basis of the present invention. Therefore, these modifications or improvements made on the basis of not departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. A method for controlling charging of a mobile energy storage device, characterized in that: include: Acquire interaction data between a mobile energy storage object user and a charging station, and use the interaction data to construct a graph model, wherein nodes in the graph model represent mobile energy storage object users or charging stations, and edges in the graph model represent interaction relationships between users or between users and charging stations; Extracting feature vectors from each node in the graph model, wherein the feature vectors of the mobile energy storage object user nodes in the graph model include historical charging behaviors and charging preference parameters; the feature vectors of the charging station nodes in the graph model include power station location, capacity, current status, and historical load parameters; Constructing a graph attention network model, which uses the feature vector extracted from each node and aggregates the information of neighboring nodes by learning the attention weights between nodes to predict the user's charging needs and the available power of the charging station; A charging station or several charging stations in a geographical area is regarded as a local node, each of the local nodes trains a graph attention network model through local data, obtains parameters of the graph attention network model of each local node, aggregates and updates the parameters of the graph attention network model of each local node to optimize the global model, and distributes the optimized global model parameters to each local node; Find the global optimal solution based on the prediction results of the optimized global model, and obtain the optimal charging station allocation and charging time plan for each mobile energy storage object; Taking a charging station or several charging stations in a geographical area as a local node, the calculation formula for each local node to train a graph attention network model through local data is: In the formula, represents the model parameters at node i in round t+1; represents the model parameters on node i in the tth round, η is the hyperparameter learning rate, Represents the gradient of the loss function calculated by node i on the local data; Before obtaining the parameters of the graph attention network model of each of the local nodes, the method further includes: For each local node, differential privacy noise is added to the model update of the local node, and the expression formula is: In the formula, represents the model parameters on node i in the t+1th round; Noise(∈) represents the added differential privacy noise; represents the model parameters after adding differential privacy noise to node i in round t+1; The expression formula for aggregating and updating the parameters of the graph attention network model of each local node to optimize the global model is: Where N represents the total number of nodes participating in federated learning; θ (t+1) Represents the parameters of the global model after t+1 rounds of iterations, which are obtained by aggregating the noisy model parameters uploaded by all participating nodes and are used to update the global model on the central server; According to the prediction results of the optimized global model, the global optimal solution is found, and the optimal charging station allocation and charging time planning process for each mobile energy storage object is obtained: Define a set of mobile energy storage objects E = {e1, e2, ..., e n In a group of charging stations S = {s1, s2, ..., s m } for charging; each mobile energy storage object has a charging demand d i and arrival time a i ; Each charging station j Maximum available power C j and a set of charging piles; the goal is to assign the most suitable charging station and charging pile to each mobile energy storage object, and plan the charging time of each mobile energy storage object to minimize the total charging time and cost while considering the grid load balance; Define a binary decision vector x ij , if the mobile energy storage object e i Assigned to s j , then x ij =1, otherwise x ij =0; Charging time t for each mobile energy storage object i By charging demand i and the charging rate r of the charging station j The calculations show that: The optimization goal is to minimize the total charging cost z, including charging time, charging waiting time and optimizing grid load balance. Assuming that the charging cost is proportional to the charging time, the objective function is expressed as: Where α and β are hyperparameter balancing coefficients, which are used to adjust the relative importance of charging time and grid load; L j Representative charging stations j The load is the total charging power of all mobile energy storage objects assigned to the station; is the average load of all charging stations, which is used to evaluate the degree of load balancing; The constraints of the global model include: Each mobile energy storage object is assigned to only one charging station: Charging station capacity limitations: Restrictions on arrival time, charging time and charging station operation time of mobile energy storage objects: The optimal charging station allocation and charging time plan for each mobile energy storage object are obtained by solving the problem; After obtaining the optimal charging station allocation and charging time plan for each mobile energy storage object, it also includes evaluating the execution effect of the current charging scheduling strategy and adjusting the current charging scheduling strategy based on the evaluation results; In the process of evaluating the execution effect of the current charging scheduling strategy, a charging waiting time evaluation model, a power grid load balance evaluation model and an energy cost evaluation model are constructed; The expression formula of the charging waiting time evaluation model is: Where W` represents the average waiting time, P represents the total number of mobile energy storage objects charged during the evaluation period, and w i represents the waiting time of the i-th vehicle; The expression formula of the power grid load balance evaluation model is: Where, L var represents the variance of the grid load; L j Representative charging stations j The load is the total charging power of all mobile energy storage objects assigned to the station; is the average load of all charging stations, which is used to evaluate the degree of load balancing; M is the total number of charging stations; The expression formula of the energy cost assessment model is: In the formula, C total is the total energy cost, c i is the unit energy cost of charging the i-th vehicle, d i is the charging demand of the i-th vehicle.

2. A mobile energy storage device charging control method according to claim 1, characterized in that: In the process of building the graph attention network model, the feature vector h′ updated for each node i is i The calculation formula is: In the formula, represents the set of neighbor nodes of node i, α ij represents the attention coefficient; W is the trainable weight matrix, h j represents the feature vector of node j, and softmax(·) represents the nonlinear activation function; Attention coefficient α ij The calculation formula is: In the formula, a T represents a trainable weight vector used to calculate the attention coefficient between any two nodes i and j; LeakyReLU(·) is a linear rectification function used to introduce nonlinear features; is the normalization factor, ensuring that the sum of the normalization coefficients of all neighbor nodes i of node i is 1; || Indicates a connection operation; In the process of building the graph attention network model, the loss function is defined The graph attention network model is trained by gradient descent method, and the loss function The calculation formula is: Where |V| represents the number of nodes in the graph attention network model, that is, the total number of mobile energy storage device users and charging stations, and y i is the true label data of node i, h′ i is the updated feature vector for each node i.

3. A mobile energy storage device charging control device, using a mobile energy storage device charging control method according to any one of claims 1 to 2, characterized in that: include: A graph model building module, used to obtain interaction data between mobile energy storage object users and charging stations, and use the interaction data to build a graph model, wherein nodes in the graph model represent mobile energy storage object users or charging stations, and edges in the graph model represent interaction relationships between users, or between users and charging stations; A feature vector extraction module is used to extract feature vectors from each node in the graph model, wherein the feature vectors of the mobile energy storage object user nodes in the graph model include historical charging behaviors and charging preference parameters; the feature vectors of the charging station nodes in the graph model include power station location, capacity, current status, and historical load parameters; A graph attention network construction module is used to construct a graph attention network model. The graph attention network model uses the feature vector extracted from each node to aggregate the information of neighboring nodes by learning the attention weights between nodes to predict the user's charging needs and the available power of the charging station; A federated learning optimization module, used to treat a charging station or several charging stations in a geographical area as a local node, each of the local nodes trains a graph attention network model through local data, obtains parameters of the graph attention network model of each local node, aggregates and updates the parameters of the graph attention network model of each local node to optimize the global model, and distributes the optimized global model parameters to each local node; The resource scheduling and allocation module is used to find the global optimal solution based on the optimized global model prediction results, and obtain the optimal charging station allocation and charging time plan for each mobile energy storage object; Also includes: The strategy evaluation and adjustment module is used to evaluate the execution effect of the current charging scheduling strategy and adjust the current charging scheduling strategy according to the evaluation results.

Images